Vision-Based Yawning Classification System for Real-Time Drowsiness Detection
Authors: Dr. Ana María Torres, Dr. Carlos Eduardo Rivas, Dr. Lucia Fernández
DOI: 10.87349/JBUPT/28304
Page No: 29-34
Abstract
As the facts depict drowsiness or fatigue is one of the major causes of road accident. Drowsiness impairs driver’s senses and the driver is not able to respond quickly resulting in higher risk of road accidents. Being a significant indicator of drowsiness, yawning detection has been explored in this paper. Many geometric and feature based methods have been previously proposed but they are sensitive to illumination, appearance variations such as skin color etc. Hence, effective feature extraction techniques are needed that extract good features in the presence of varying lightning conditions and further improve classification accuracy. Frequency domain feature extraction has not been used extensively in this area. In this paper, Local energy based shape histogram (LESH) has been implemented that extracts the features in the frequency domain, as a result it is insensitive to contrast and illumination. Support Vector machine (SVM) is a widely used machine learning tool because of its binary nature of classification and also efficiently classifying non-linear data. The proposed framework is evaluated on Yawdd database. A sensitivity of 100%, specificity of 94% and accuracy of 95% was achieved.
